The invention applies advantageously in the case of a quantity that varies continuously, that is to say the value whereof changes continuously over time—for example, the value whereof changes at least once per second.
It also relates to a computer program and an associated device, along with a method for monitoring quantities of computer devices.
In the present description, and in the claims, the following terminology is used.
The present invention uses a sequence of values of a quantity, that is to say successive values adopted by the quantity at discrete times, equally distributed over time or not.
A difference between two values indicates a difference in value and not a timing difference.
A value preceding another value is the value situated immediately before this other value in the sequence of values (there is no value between them).
A value following another value is the value situated immediately after this other value in the sequence of values (there is no value between them).
A prior value in respect of another value is a value situated before this other value in the sequence of values (there may be one or a plurality of values between them).
A posterior value in respect of another value is a value situated after this other value in the sequence of values (there may be one or a plurality of values between them).
A local extremum is a value either greater or less than both the value preceding same and the value following same.
A small variation value is a posterior value in respect of a local extremum transmitted, referred to as the last local extremum before small variation, having therewith a difference less than a predefined difference and which either follows same, or is only separated therefrom by one or a plurality of other small variation values, that is to say which each have a difference less than the predefined difference therewith.
A last small variation value is a small variation value which is followed by a measured value having a difference with the last local extremum before small variation greater than the predefined difference.
Most computer device monitoring applications still currently use a regular querying method triggered by an administration server, to obtain information from the computer devices monitored.
When such an application monitors a small number of computer devices, this method has a number of advantages, but when the number of computer devices monitored increases for example in the case of large sets of computer devices (referred to as “clusters”) for intensive computing, this method very soon reaches the limits thereof: indeed, the queries sent by the administration server increase in a linear fashion with the number of computer devices monitored and lead to saturation of the pass-band of the administration network or the power available on the administration server.
A conventional response to this situation is that of switching to “push” mode monitoring: the computer devices monitored take the initiative to send an item of information relating to said device to the administration server, which remains continuously on standby for these notifications. This approach is perfectly suitable for reporting “accidental” or occasional events, such as faults, temperature threshold overshoots. Indeed, these events are low in frequency and replacing a frequently unnecessary regular query sent by the administration server by the transmission from the computer device in question of the relevant information solely at the time of the occurrence of the event makes it possible to reduce the need for network pass-band and processing power on the administration server to the minimum thereof.
This conventional response is not as useful if the monitoring does not relate to events with a low frequency, but to a continuous quantity of the computer device: the electrical consumption thereof for example.
In the case of a continuous quantity, it is known to define a certain number of value thresholds, regularly spaced or not and covering the entire range of possible values for the quantity. Each overshoot of a threshold is then considered to be an event, giving rise to a transmission of the value of the quantity by the computer device. This technique has the drawback of not being precise when the values of the quantity stabilize and change between two thresholds.
For this purpose, a method for monitoring a quantity of a computer device is proposed, comprising:                measuring values adopted by the quantity over time,        
characterized in that it further comprises:                determining a measured value, or extremum value, meeting at least one transmission criterion, the transmission criterion/criteria met by the extremum value comprising: the extremum value is a local extremum,        transmitting the extremum value.        
Optionally, the transmission criterion/criteria met by the extremum value further comprise: the extremum value has a difference with the last local extremum transmitted greater than a predefined difference.
Indeed, it has been observed that a transmitted local extremum is frequently followed by a certain number of values varying little relative to said value, for example oscillating around the local extremum value. Transmitting all the local extrema during these small variations would merely be of little interest since they would all have almost the same value. In this way, adding the transmission criterion whereby the extremum value has a difference with the last local extremum transmitted greater than a predefined difference makes it possible to prevent transmission during small variations.
Optionally, the last local extremum transmitted, or last local extremum before small variation, is followed by one or a plurality of measured values, or small variation values, each having a difference with the last local extremum before small variation, less than the predefined difference.
Optionally, the method further comprises:                determining the last small variation value, that is to say the small variation value which is followed by a measured value having a difference with the last local extremum before small variation greater than the predefined difference,        transmitting the last small variation value.        
By means of the transmission of the last small variation value, the administration server knows how long the small variations lasted, and thus has a more accurate idea of the progression of the monitored quantity.
Optionally, the method further comprises:                determining a measured value, or crossing value, such that one of the crossing value and the measured value preceding same is greater than a predefined threshold whereas the other value is less than this predefined threshold,        transmitting the crossing value.        
Optionally, the measured values other than the extremum values and/or the last small variation values and/or the crossing values are not transmitted.
Optionally, the monitored quantity is the electrical consumption of the computer device.
A method for monitoring quantities of computer devices is also proposed, each computer device being connected to the same administration server via a communication network, comprising: for each computer device, performing a method for monitoring a quantity of the computer device according to the invention, wherein the transmitted value(s) are transmitted to the administration server via the communication network.
A computer program downloadable from a communication network and/or saved on a computer-readable medium and/or executable by a processor is also proposed, characterized in that it comprises instructions for executing steps of a monitoring method according to the invention, when said program is executed on at least one computer.
A computer device for monitoring a quantity of a computer device is also proposed, comprising:                means for measuring values adopted by the quantity over time, characterized in that it further comprises:        means for determining a measured value, or extremum value, meeting at least one transmission criterion, the transmission criterion/criteria met by the extremum value comprising: the extremum value is a local extremum,        means for transmitting the extremum value.        